Abstract: The conditional nonlinear optimal perturbations (CNOPs) obtained by a fast algorithm are applied to the targeting observations of Typhoon Matsa (2005) based on the Global/Regional Assimilation and PrEdiction System (GRAPES), an operational regional prediction model. Through a series of sensitivity experiments, several issues on targeting strategy design are discussed, including the effectivity of different guidance to determine the sensitive area (or targeting area) and the impact of the size of targeting area on the improvement of 24hour forecast. Three guidances are used when employing the CNOP for finding the targeting area to improve the 24hour sea lever pressure prediction and 24hour accumulated rainfall prediction in the verification region (i.e. 116.8125°—129.75°E, 20.125°—35.3125°N). The distribution and effectivity of these targeting areas are compared with each other with the results that the sensitive areas identified by the three guidances used in this study are different from each other. All the sensitive areas determined by these guidances are effective in that the improvement of the 24hour forecast is reached. The last two guidances are more effective and able to identify more similar sensitive areas than the first one. Further, when the size of targeting areas is enlarged twice it is noticed that the 24hour accumulated rainfall prediction is not necessarily improved since although an initial condition improvement in a larger size of targeting area may sometimes result in a better prediction skill, it is not always true, depending on the sensitivity of the guidance to the size of targeting area.